Automatic emergency braking decision method and automatic emergency braking control method

Abstract

The application provides an automatic emergency brake decision method and an automatic emergency brake control method, relates to the field of vehicle active safety technology, and completely breaks through the information attenuation bottleneck caused by the series of a perception layer and a decision layer in a traditional scheme by introducing a first automatic emergency brake model, a second automatic emergency brake model and a human rule decision model, and taking original sensor data and vehicle state data as the input of the first automatic emergency brake model, and eliminates the single-point failure hidden danger existing in a single model. Moreover, the method avoids the sensitivity adjustment problem caused by the fact that a fixed threshold is used as the basis for triggering decision in a traditional scheme through comprehensive decision, can not only guarantee the efficient and sensitive response to complex and changeable actual road conditions, but also can improve the robustness and reliability of the active safety system of the target vehicle in all directions through avoiding mis-triggering behaviors such as ghost braking, and reduce the probability of vehicle accidents.

Description

Technical Field

[0001] This invention relates to the field of vehicle active safety technology, and in particular to an automatic emergency braking decision-making method and an automatic emergency braking control method. Background Technology

[0002] With the continuous development of advanced driver assistance systems (ADAS), automatic emergency braking (AEB), as a core function of vehicle active safety, is becoming increasingly important. The goal of the AEB system is to reduce or completely avoid collisions by automatically braking when the driver's reaction is insufficient or their attention is distracted.

[0003] In terms of technical implementation, existing AEB (Autonomous Emergency Braking) algorithm architectures typically consist of three parts: a perception layer, a decision layer, and an execution layer. The perception layer primarily utilizes multi-sensor fusion technology, including millimeter-wave radar, ultrasonic radar, lidar, and cameras, to acquire environmental perception data. The decision layer continuously calculates the relative speed, distance, and time to collision (TTC) between the vehicle and obstacles ahead based on the environmental perception data. It then combines this with a collision avoidance strategy based on safe distance or a collision time-based avoidance strategy to conduct risk assessment and trigger braking commands. The execution layer, based on the braking commands triggered by the decision layer, completes the braking action through Electronic Hydraulic Brake (EHB) or Brake-by-Wire (BBW) systems. In specific situations, it may also coordinate with Automatic Emergency Steering (AES) to jointly avoid collisions.

[0004] However, existing AEB algorithms suffer from significant information loss when transmitting information from the perception layer to the decision layer due to issues such as the high dependence of the output of the perception layer on a whitelist learned under supervision based on specific target categories and the sparsity of the occupancy grid (OCC). Furthermore, using a fixed threshold as the trigger for decisions presents a dilemma in sensitivity adjustment: an overly conservative strategy can lead to delayed braking or missed triggers, while an overly aggressive strategy can misjudge common objects, frequently triggering phantom braking, severely interfering with normal driving and even causing rear-end collisions. Summary of the Invention

[0005] This invention provides an automatic emergency braking decision-making method and an automatic emergency braking control method to address the deficiencies in related technologies.

[0006] This invention provides an automatic emergency braking decision-making method, applied to a target vehicle, comprising: Acquire sensor data and vehicle status data of the target vehicle; The sensor data and the vehicle status data are input into the first automatic emergency braking model to obtain the environmental perception data and the first collision dynamics parameters output by the first automatic emergency braking model. The environmental perception data and the vehicle status data are input into the second automatic emergency braking model to obtain the second collision dynamics parameters output by the second automatic emergency braking model. The environmental perception data and the vehicle state data are input into the human rule decision model to obtain the third collision dynamics parameter output by the human rule decision model; The automatic emergency braking decision is determined by comprehensively considering the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter.

[0007] According to an automatic emergency braking decision-making method provided by the present invention, the step of determining the automatic emergency braking decision result by making a comprehensive decision based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter includes: Based on the first collision dynamics parameters, the risk level of the current scenario of the target vehicle is classified. If the result obtained from the classification is the target level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the first collision dynamics parameter is verified to determine the automatic emergency braking decision result; If the result of the classification is not the target level, then the automatic emergency braking decision result is determined based on the first collision dynamics parameter.

[0008] According to an automatic emergency braking decision-making method provided by the present invention, the target level includes a warning level and an emergency level; the first collision dynamics parameter includes collision time, braking deceleration and steering angle. The step of verifying the first collision dynamics parameter based on the second and third collision dynamics parameters to determine the automatic emergency braking decision result includes: If the result obtained by the grading is the warning level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the collision time and braking deceleration in the first collision dynamics parameter are verified to determine the automatic emergency braking decision result; If the result obtained from the classification is the emergency level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the collision time, braking deceleration and steering angle in the first collision dynamics parameter are verified to determine the automatic emergency braking decision result.

[0009] According to an automatic emergency braking decision method provided by the present invention, the step of verifying the first collision dynamics parameter based on the second collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result further includes: Calculate the first decision deviation between the first collision dynamics parameter and the second collision dynamics parameter, and the second decision deviation between the first collision dynamics parameter and the third collision dynamics parameter, and determine the automatic emergency braking decision result based on the first decision deviation and / or the second decision deviation.

[0010] According to an automatic emergency braking decision-making method provided by the present invention, the step of calculating a first decision deviation between a first collision dynamics parameter and a second collision dynamics parameter, and a second decision deviation between the first collision dynamics parameter and the third collision dynamics parameter, and determining the automatic emergency braking decision result based on the first decision deviation and / or the second decision deviation, includes: Calculate the first decision bias; If the first decision deviation is less than the first preset threshold, then the first collision dynamics parameter is determined as the automatic emergency braking decision result; If the first decision deviation is greater than or equal to the first preset threshold, then the second decision deviation is calculated; If the second decision deviation is less than the second preset threshold, the first collision dynamics parameter is determined as the automatic emergency braking decision result; if the second decision deviation is greater than or equal to the second preset threshold, the automatic emergency braking decision result is determined to be empty.

[0011] According to an automatic emergency braking decision-making method provided by the present invention, the step of determining the automatic emergency braking decision result by making a comprehensive decision based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter further includes: The operational status of the first automatic emergency braking model and the second automatic emergency braking model is evaluated in real time. If both the first automatic emergency braking model and the second automatic emergency braking model are operating normally, a comprehensive decision is made based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter to determine the automatic emergency braking decision result.

[0012] According to an automatic emergency braking decision-making method provided by the present invention, the method further includes, after real-time evaluation of the operational states of the first automatic emergency braking model and the second automatic emergency braking model: If the first automatic emergency braking model is operating normally and the second automatic emergency braking model is operating abnormally, a comprehensive decision is made based on the first collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result. If the first automatic emergency braking model is in an abnormal operating state while the second automatic emergency braking model is in a normal operating state, a comprehensive decision is made based on the second collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result. If the operational states of both the first automatic emergency braking model and the second automatic emergency braking model are abnormal, the automatic emergency braking decision result is determined based on the third collision dynamics parameter.

[0013] According to an automatic emergency braking decision-making method provided by the present invention, the training steps of the first automatic emergency braking model include: The sensor data samples, vehicle state samples, and actual collision dynamics parameters of the sample vehicle are obtained, and the sensor data samples are labeled with environmental perception data tags. The sensor data samples and the vehicle state samples are input into the first initial model to obtain the environmental perception results and the first collision dynamics parameter prediction results output by the first initial model. Based on the difference between the environmental perception result and the environmental perception data label, and the difference between the first collision dynamics parameter prediction result and the actual collision dynamics parameter, the first initial model is iteratively trained to obtain the first automatic emergency braking model.

[0014] According to an automatic emergency braking decision-making method provided by the present invention, the training steps of the second automatic emergency braking model include: Acquire environmental perception data samples, vehicle state samples, and actual collision dynamics parameters of the sample vehicle; The environmental perception data samples and the vehicle state samples are converted into data vectors, and the data vectors are input into the second initial model to obtain the second collision dynamics parameter prediction results output by the second initial model; Based on the difference between the predicted collision dynamics parameters and the actual collision dynamics parameters, the second initial model is iteratively trained to obtain the second automatic emergency braking model.

[0015] The present invention also provides an automatic emergency braking control method, applied to a target vehicle, comprising: The automatic emergency braking decision result of the target vehicle is obtained, and the automatic emergency braking decision result is obtained based on the automatic emergency braking decision method described above. Based on the automatic emergency braking decision results, automatic emergency braking control is performed on the target vehicle.

[0016] The present invention also provides an automatic emergency braking decision device, applied to a target vehicle, comprising: The data acquisition module is used to acquire sensor data and vehicle status data of the target vehicle; The first decision module is used to input the sensor data and the vehicle status data into the first automatic emergency braking model to obtain the environmental perception data and the first collision dynamics parameters output by the first automatic emergency braking model. The second decision module is used to input the environmental perception data and the vehicle status data into the second automatic emergency braking model to obtain the second collision dynamics parameters output by the second automatic emergency braking model. The third decision module is used to input the environmental perception data and the vehicle state data into the human rule decision model to obtain the third collision dynamics parameters output by the human rule decision model. The integrated decision module is used to make an integrated decision based on the first collision dynamics parameter, the second collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result.

[0017] The present invention also provides an automatic emergency braking control device, applied to a target vehicle, comprising: The decision result acquisition module is used to acquire the automatic emergency braking decision result of the target vehicle, which is obtained based on the automatic emergency braking decision method described above. The brake control module is used to perform automatic emergency braking control on the target vehicle based on the automatic emergency braking decision result.

[0018] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the automatic emergency braking decision method or the automatic emergency braking control method as described above.

[0019] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the automatic emergency braking decision method or the automatic emergency braking control method as described above.

[0020] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the automatic emergency braking decision method or the automatic emergency braking control method as described above.

[0021] The automatic emergency braking decision-making method and automatic emergency braking control method provided by this invention introduce a first automatic emergency braking model, a second automatic emergency braking model, and a human-rule-based decision-making model. By using raw sensor data and vehicle state data as input to the first automatic emergency braking model, this method completely overcomes the information attenuation bottleneck caused by the serial connection between the perception layer and the decision-making layer in traditional solutions, eliminating the single-point failure risk inherent in individual models. Furthermore, through comprehensive decision-making, this method avoids the sensitivity adjustment problem caused by using fixed thresholds as the triggering basis in traditional solutions. This not only ensures efficient and sensitive response to complex and changing real-world road conditions but also comprehensively improves the robustness and reliability of the target vehicle's active safety system by avoiding false triggering behaviors such as phantom braking, thereby reducing the probability of vehicle accidents. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in this invention or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the automatic emergency braking decision-making method provided by the present invention.

[0024] Figure 2 This is a schematic diagram of the steering angle of the target vehicle to avoid colliding with an obstacle in the automatic emergency braking decision method provided by the present invention.

[0025] Figure 3 This is a flowchart illustrating the automatic emergency braking control method provided by the present invention.

[0026] Figure 4 This is a flowchart illustrating the automatic emergency braking decision-making method and the automatic emergency braking control method provided by the present invention.

[0027] Figure 5 This is a schematic diagram of the automatic emergency braking decision device provided by the present invention.

[0028] Figure 6 This is a schematic diagram of the automatic emergency braking control device provided by the present invention.

[0029] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0031] Current AEB (Autonomous Emergency Braking) algorithms use sensors such as millimeter-wave radar, ultrasonic radar, lidar, and cameras to detect road conditions and target information ahead, and calculate the collision risk level in real time. When the calculated collision risk level reaches the critical warning point, it indicates a possibility of collision with the target ahead. A warning is first issued to the driver through various sound and image methods, reminding the driver to take evasive action. Simultaneously, the brake fluid pressure is pre-filled to achieve a faster and higher braking level when braking is required. If the driver does not react correctly to the warning, partial braking is required, and additional warnings are issued through slight vibrations of the brake pedal or steering wheel. When the calculated collision risk level reaches the critical braking point, it indicates that a collision with the obstacle is almost unavoidable, and full braking is required for collision avoidance.

[0032] In the existing AEB algorithm architecture, the perception layer mainly utilizes multi-sensor fusion technology, including millimeter-wave radar, ultrasonic radar, lidar, and cameras, to acquire environmental perception data. Millimeter-wave radar, with its strong adaptability to adverse weather conditions, can detect the distance, speed, and orientation of obstacles ahead; cameras can identify rich information such as lane lines, traffic signs, and obstacles; lidar constructs a 3D environmental model with higher resolution; and ultrasonic radar has superior close-range detection capabilities, supplementing visual perception. These sensors transmit the obtained sensor data to the onboard controller in real time, providing a basis for decision-making.

[0033] The decision-making layer continuously calculates the relative speed, distance, and collision time between the vehicle and obstacles ahead based on environmental perception data. It then combines this with a collision avoidance strategy based on safe distance or a collision time-based strategy to conduct a risk assessment and trigger a braking command. A braking command is triggered if the collision time is below a safe threshold and the driver does not respond. To avoid false triggers, driver behavior, such as steering wheel movement and accelerator pedal status, is also considered for optimized judgment.

[0034] The execution layer, based on the braking command triggered by the decision-making layer, completes the braking action through the electro-hydraulic braking system or the brake-by-wire system, rapidly establishing braking pressure to decelerate or stop the vehicle. In specific situations, it will also coordinate with the automatic emergency steering function to avoid collisions.

[0035] The decision-making algorithms adopted by the decision-making layer mainly include collision avoidance strategies based on safe distance, collision avoidance strategies based on collision time, and braking-steering cooperative collision avoidance strategies.

[0036] Collision avoidance strategy based on safe distance refers to using the real-time distance between the vehicle and the vehicle in front as an indicator of driving hazard, comparing it with the warning safe distance and the emergency braking safe distance, and then making warning and active braking collision avoidance operations. The warning safe distance and the emergency braking safe distance are mainly determined in real time by parameters such as the vehicle's speed, the relative speed between the vehicle and the vehicle in front, the vehicle's maximum braking deceleration, the target vehicle's maximum braking deceleration, system delay time, driver reaction time, and the distance to be maintained after the vehicles stop. For example, the emergency braking safe distance can be calculated using the following formula: ; in, For emergency braking safety distance, For the vehicle's speed, The relative speed between the vehicle and the vehicle in front is given. , The speed of the obstacle, This is the maximum braking deceleration of the vehicle. This is the maximum braking deceleration of the vehicle in front. For system delay time, For driver reaction time, This refers to the distance that you need to maintain from the vehicle in front of you after parking.

[0037] Collision avoidance strategies based on time-of-collision (TTC) measure the remaining time between the vehicle's current motion and the collision with an obstacle, which better aligns with the driver's perception and judgment of potential hazards. This strategy compares the real-time calculated TTC with pre-set collision warning TTC and emergency braking TTC to determine appropriate collision avoidance actions, including warnings and automatic braking. For TTC calculation, the acceleration of both the vehicle and the obstacle can also be considered, employing Extended Time to Collision (ETTC).

[0038] The formulas for calculating TTC and ETTC can be expressed as follows: ; in, The distance between the vehicle and the obstacle. The speed of the obstacle, This is the difference between the vehicle's acceleration and the obstacle's acceleration.

[0039] The braking and steering-integrated collision avoidance strategy considers avoiding collisions through steering when braking cannot prevent them. It uses the AEB function in conjunction with the Automatic Emergency Steering (AES) function to avoid collisions. The AEB function focuses on longitudinal deceleration / brake, while the AES function focuses on lateral active obstacle avoidance.

[0040] However, while existing AEB algorithms can improve driving safety, they suffer from significant drawbacks in information transmission between the perception layer, decision layer, and between the perception and decision layers. The perception layer's output heavily relies on a whitelist learned under supervised learning based on specific target categories. This whitelist includes obstacle types such as vehicles, pedestrians, and two-wheeled vehicles, as well as ground texture information like lane lines, edges, and drivable areas, but it cannot identify targets outside the whitelist. Furthermore, when using an occupancy grid (OCC) method to identify targets, the occupancy grid suffers from sparsity. These factors lead to severe information loss when the perception layer transmits information to the decision layer. Additionally, using a fixed threshold as the trigger for decisions makes it difficult to dynamically adapt to complex scenarios. When environmental changes exceed the fixed threshold, it may result in delayed braking due to a conservative strategy or false triggering due to an incomplete strategy. Furthermore, the contradiction between false triggering and missed triggering leads to a dilemma in sensitivity adjustment: an overly conservative strategy may result in braking too late or missed triggering, reducing system reliability, while an overly aggressive strategy may lead to misjudgment of common objects and frequent ghost braking, seriously interfering with normal driving and even causing rear-end collisions.

[0041] Based on this, the present invention provides an automatic emergency braking decision-making method, such as... Figure 1 As shown, this method is applied to the target vehicle and includes: S11, acquire the sensor data and vehicle status data of the target vehicle; S12, input the sensor data and the vehicle status data into the first automatic emergency braking model to obtain the environmental perception data and the first collision dynamics parameters output by the first automatic emergency braking model; S13, input the environmental perception data and the vehicle status data into the second automatic emergency braking model to obtain the second collision dynamics parameters output by the second automatic emergency braking model; S14, input the environmental perception data and the vehicle state data into the human rule decision model to obtain the third collision dynamics parameter output by the human rule decision model; S15, based on the first collision dynamics parameter, the second collision dynamics parameter and the third collision dynamics parameter, a comprehensive decision is made to determine the automatic emergency braking decision result.

[0042] Specifically, the automatic emergency braking decision-making method provided in this embodiment of the invention is executed by an automatic emergency braking decision-making device, which can be configured in the target vehicle's onboard controller, autonomous driving domain controller, onboard computer, or cloud server. The target vehicle is the vehicle currently in motion that requires collision risk assessment and intervention.

[0043] The target vehicle may be equipped with various sensors, either externally or internally, such as millimeter-wave radar, ultrasonic radar, lidar, and cameras, to detect various environmental data, including the distance, speed, and orientation of obstacles ahead, lane lines, traffic signs, obstacles, and 3D environmental models. The cameras can be monocular, binocular, panoramic, or surround-view cameras, etc., without specific limitations.

[0044] The target vehicle can be equipped with an Inertial Measurement Unit (IMU) and a positioning system. The IMU measures the target vehicle's attitude data, and the positioning system locates the target vehicle, determining its position data. The target vehicle's attitude data and position data constitute its pose data.

[0045] First, step S11 is executed to acquire sensor data from the target vehicle. This sensor data may include environmental data detected by various sensors installed on the target vehicle, and vehicle status data may include at least one of the following: the target vehicle's speed, acceleration, steering wheel angle, and pose data.

[0046] By directly acquiring abundant and unprocessed raw multimodal sensor data and vehicle status data, the most comprehensive and fundamental scene representation can be provided for the subsequent first automatic emergency braking model, thus avoiding the problem of loss of key environmental information caused by the limitations of the early perception whitelist in traditional solutions.

[0047] Then, step S12 is executed, inputting sensor data and vehicle state data into the first automatic emergency braking model to obtain environmental perception data and first collision dynamics parameters output by the first automatic emergency braking model. The first automatic emergency braking model is an end-to-end neural network model built based on deep learning, which may include a state data encoding module, an image feature extraction module, and a perception module. The state data encoding module encodes the vehicle state data to obtain state codes. The image feature extraction module extracts features from the images in the sensor data to obtain image features. The perception module uses the state codes and image features to generate environmental perception data and the first collision dynamics parameters. The environmental perception data may include obstacle detection results, lane line detection results, and occupancy grid detection results. The first collision dynamics parameters may include braking deceleration, collision time, and collision avoidance steering angle.

[0048] Braking deceleration refers to the deceleration parameter required for the longitudinal motion control of the target vehicle to avoid a collision or mitigate collision damage; collision time refers to the remaining time scale from the current moment until the target vehicle collides with the obstacle in front while moving in its current state; steering angle refers to the front wheel angle or steering control parameter required for the lateral motion control of the target vehicle in conjunction with the emergency steering function to avoid a collision.

[0049] In this step, by directly inputting the raw sensor data into the first automatic emergency braking model, the barrier between the traditional perception layer and the decision-making layer can be broken down. This solves the pain point that key targets cannot enter the risk assessment stage due to the restriction of the perception whitelist, and realizes high real-time basic decision-making with zero information loss, which greatly improves the model's generalization and adaptability to complex road scenarios.

[0050] Next, step S13 is executed, introducing a second automatic emergency braking model. Environmental perception data and vehicle state data are input into the second automatic emergency braking model, and the second collision dynamics parameters are obtained through the second automatic emergency braking model. The second automatic emergency braking model refers to a heterogeneous end-to-end neural network model based on deep learning, which exists as an independent redundant computing layer. It does not directly process the original sensor data, but instead performs calculations on the environmental perception data obtained through the first automatic emergency braking model to obtain the second collision dynamics parameters.

[0051] The second automatic emergency braking model may include a state data encoding module, a vector encoding module, and a perception module. The state data encoding module encodes vehicle state data to obtain state codes. The vector encoding module performs feature encoding on environmental perception data to obtain vector encoded features. The perception module uses the state codes and vector encoded features to generate second collision dynamics parameters. These second collision dynamics parameters are physical quantities of the same type as the first collision dynamics parameters, independently calculated by the second automatic emergency braking model, including braking deceleration, collision time, and steering angle for collision avoidance, used for subsequent comprehensive decision-making.

[0052] The second automatic emergency braking model, based on the reuse of environmental perception data, can realize a redundancy verification mechanism for homogeneous data and heterogeneous computing. Without increasing the cost of additional sensor hardware, it can effectively prevent the logical blind spots and single point of failure risks that may exist in a single model, and can also avoid the loopholes that may exist in manually defined rules.

[0053] Next, step S14 is executed, introducing a human-rule decision-making model. Environmental perception data and vehicle state data are input into the human-rule decision-making model, and the third collision dynamics parameters are obtained through the output of the human-rule decision-making model. The human-rule decision-making model refers to a human-machine collaborative verification layer algorithm model built on traditional physics, kinematic formulas, and prior human driving logic, without relying on the black-box mechanism of neural networks. It filters out obstacles with the greatest impact on the target vehicle collision through human-preset logical rules, and strictly calculates the third collision dynamics parameters based on the geometric relationship characteristics such as the target vehicle's width and longitudinal distance from the target, combined with pose data such as relative velocity, using preset physical kinematic formulas.

[0054] The third collision dynamics parameter refers to a physical quantity of the same type as the first collision dynamics parameter, which is independently calculated by the human rule decision model. It includes braking deceleration, collision time, and steering angle to avoid collision, and is used for subsequent comprehensive decision-making.

[0055] For example, the collision time can be calculated using the following formula: .

[0056] like Figure 2 As shown, the turning radius R of the target vehicle to avoid colliding with the obstacle can be expressed as: ; Where L is the longitudinal distance between the target vehicle and the obstacle, and w is the width of the vehicle.

[0057] The target vehicle's steering angle to avoid colliding with obstacles It can be calculated using the following formula: .

[0058] This step introduces a human-rule decision-making model with strong interpretability and determinism as a safety net verification layer, which can effectively avoid unpredictable extreme errors that may occur in long-tail edge scenarios by deep learning black-box models such as the first automatic emergency braking model and the second automatic emergency braking model.

[0059] Finally, step S15 is executed, where a comprehensive decision is made using the first, second, and third collision dynamics parameters to determine the automatic emergency braking decision. The comprehensive decision-making process can be a holistic macro-evaluation process utilizing the first, second, and third collision dynamics parameters, including weight allocation, cross-validation, state degradation, and scenario classification. The automatic emergency braking decision is the final vehicle-level instruction, encompassing information such as whether automatic emergency braking is needed, whether longitudinal braking or lateral steering should be prioritized, and the specific intervention control amount. Therefore, the automatic emergency braking decision can be empty (i.e., no information is output), or it can include braking deceleration, collision time, and the steering angle for collision avoidance.

[0060] If the automatic emergency braking decision result is not empty, it can be obtained by weighted summation of similar parameters among the first, second, and third collision dynamics parameters. Alternatively, the automatic emergency braking decision result can be obtained by selecting the maximum braking deceleration, the maximum collision time, and the maximum steering angle among similar parameters among the first, second, and third collision dynamics parameters.

[0061] The automatic emergency braking decision-making method provided in this embodiment of the invention introduces a first automatic emergency braking model, a second automatic emergency braking model, and a human-rule-based decision-making model. It uses raw sensor data and vehicle state data as input to the first automatic emergency braking model, completely overcoming the information attenuation bottleneck caused by the serial connection between the perception and decision layers in traditional solutions, and eliminating the single-point-of-failure risk inherent in individual models. Furthermore, this method, through comprehensive decision-making, avoids the sensitivity adjustment problem caused by using fixed thresholds as the triggering basis in traditional solutions. This not only ensures efficient and sensitive response to complex and changing real-world road conditions but also comprehensively improves the robustness and reliability of the target vehicle's active safety system by avoiding false triggering behaviors such as phantom braking, thereby reducing the probability of vehicle accidents.

[0062] Based on the above embodiments, the step of making a comprehensive decision based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter to determine the automatic emergency braking decision result includes: Based on the first collision dynamics parameters, the risk level of the current scenario of the target vehicle is classified. If the result obtained from the classification is the target level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the first collision dynamics parameter is verified to determine the automatic emergency braking decision result; If the result of the classification is not the target level, then the automatic emergency braking decision result is determined based on the first collision dynamics parameter.

[0063] Specifically, in the process of making comprehensive decisions, the first collision dynamics parameters output by the first automatic emergency braking model can be used as the main information, and the risk level of the target vehicle's current scenario can be classified by the first collision dynamics parameters.

[0064] Here, risk level classification can be achieved by comparing the collision time in the first collision dynamics parameter with a time threshold. Multiple time thresholds can be included, allowing for the setting of multiple risk levels. For example, the time thresholds can include a first time threshold and a second time threshold, with the first time threshold being greater than the second time threshold, thus setting three risk levels. If the collision time is greater than or equal to the first time threshold, it is a low-risk level; if the collision time is less than the first time threshold but greater than or equal to the second time threshold, it is a warning level; and if the collision time is less than the second time threshold, it is an emergency level. The warning level refers to a scenario with a moderate collision risk, such as a driving scenario where the calculated collision time is between two and four seconds. The emergency level refers to a scenario with an extremely high collision risk and an imminent collision, such as a driving scenario where the calculated collision time is less than two seconds.

[0065] The risk level increases sequentially from low risk to warning to emergency, with warning level being medium risk and emergency level being high risk. The specific values ​​of the first and second time thresholds can be set as needed; the first time threshold can be set to 4 seconds, and the second time threshold can be set to 2 seconds.

[0066] If the result of the classification is the target level, which can be the second or third level, it indicates that there is a possibility of braking. At this time, the second and third collision dynamics parameters can be used to verify the first collision dynamics parameter and determine the automatic emergency braking decision result.

[0067] For example, the decision biases between the second and third collision dynamics parameters and the first collision dynamics parameter can be calculated separately. Here, the decision bias can be the bias between parameters of the same type, which can be an absolute bias, a relative bias, or a weighted sum of the relative differences between all parameters of the same type.

[0068] If all decision deviations are less than the corresponding preset thresholds, the verification can be considered successful, and the first collision dynamics parameter can be used as the automatic emergency braking decision result.

[0069] If both decision deviations are greater than their corresponding preset thresholds, the verification is considered to have failed. In this case, either the second or third collision dynamics parameter is selected as the automatic emergency braking decision result, or the automatic emergency braking decision result is set to empty to prevent false triggering. If one decision deviation is greater than its corresponding preset threshold and the other is less than its corresponding preset threshold, the collision dynamics parameter corresponding to the decision deviation less than its corresponding preset threshold is used as the automatic emergency braking decision result.

[0070] If the result of the classification is not the target level, i.e., the first level, it means that there is no possibility of braking. In this case, the first collision dynamics parameter can be directly used as the result of the automatic emergency braking decision.

[0071] In this embodiment of the invention, the risk level of the target vehicle's current scenario is classified based on the first collision dynamics parameter without information loss. Under the target level, a comprehensive verification is performed using the second and third collision dynamics parameters. By constructing a multi-layer decision redundancy mechanism, the decision-making efficiency in low-risk scenarios is ensured while effectively overcoming the potential faults and vulnerabilities of a single sensor or single-point algorithm model. This fundamentally prevents ghost braking or false triggering behavior caused by model uncertainty, thereby improving the robustness and safety of automatic emergency braking decision-making.

[0072] Based on the above embodiments, the target level includes a warning level and an emergency level; the first collision dynamics parameter includes collision time, braking deceleration, and steering angle; The step of verifying the first collision dynamics parameter based on the second and third collision dynamics parameters to determine the automatic emergency braking decision result includes: If the result obtained by the grading is the warning level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the collision time and braking deceleration in the first collision dynamics parameter are verified to determine the automatic emergency braking decision result; If the result obtained from the classification is the emergency level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the collision time, braking deceleration and steering angle in the first collision dynamics parameter are verified to determine the automatic emergency braking decision result.

[0073] Specifically, in the process of verifying the first collision dynamics parameter using the second and third collision dynamics parameters, if the result of the classification is a warning level, it is expected that the target vehicle can avoid the collision by braking longitudinally. In order to prevent unnecessary lateral steering actions that could cause the vehicle to roll over, multiple comparisons and verifications can be performed only on the collision time and braking deceleration in the first collision dynamics parameter at the warning level.

[0074] In extremely high-risk emergency scenarios, the target vehicle may be unable to brake at all, necessitating the combined use of lateral active obstacle avoidance strategies. When activating the steering-to-collision-avoidance strategy, there is a possibility of rollover; therefore, it is necessary to rigorously verify the three key dynamic parameters in the first collision dynamics parameters—collision time, braking deceleration, and steering angle—and determine the automatic emergency braking decision only when the verifications are consistent.

[0075] In this embodiment of the invention, different verification dimensions are adopted according to the different risk levels of the warning level and the emergency level. The warning level focuses on verifying the collision time and braking deceleration, while the emergency level jointly verifies the collision time, braking deceleration and steering angle. This can dynamically adapt to the collision avoidance requirements of complex scenarios and perfectly realize the safety protection logic of prioritizing braking and then considering steering. It effectively reduces the probability of missed triggers caused by conservative strategies and maximizes the safety of the target vehicle's own driving posture while avoiding collisions.

[0076] Based on the above embodiments, the step of verifying the first collision dynamics parameter based on the second collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result further includes: Calculate the first decision deviation between the first collision dynamics parameter and the second collision dynamics parameter, and the second decision deviation between the first collision dynamics parameter and the third collision dynamics parameter, and determine the automatic emergency braking decision result based on the first decision deviation and / or the second decision deviation.

[0077] Specifically, in the process of verifying the first collision dynamics parameters using the second and third collision dynamics parameters, a first decision deviation between the first and second collision dynamics parameters, and a second decision deviation between the first and third collision dynamics parameters, can be calculated separately. The first and second decision deviations can be calculated sequentially, simultaneously, or under certain conditions; no specific limitations are imposed here. It can be understood that the first decision deviation can be the deviation between the first and second collision dynamics parameters of the same type, such as the deviation between collision times, braking decelerations, or steering angles, or it can be the weighted sum of the relative deviations between all similar parameters.

[0078] Subsequently, the automatic emergency braking decision result is determined using at least one of the first decision deviation and the second decision deviation. For example, when the first decision deviation is less than the corresponding preset threshold, the first collision dynamics parameter can be directly used as the automatic emergency braking decision result; when the second decision deviation is less than the corresponding preset threshold, the second collision dynamics parameter can be directly used as the automatic emergency braking decision result; and when both the first and second decision deviations are less than the corresponding preset thresholds, the first collision dynamics parameter can be directly used as the automatic emergency braking decision result.

[0079] In this embodiment of the invention, by calculating the first decision deviation between the first collision dynamics parameter and the second collision dynamics parameter, and the second decision deviation between the first collision dynamics parameter and the third collision dynamics parameter, and determining the final decision result based on the first decision deviation and / or the second decision deviation, multi-dimensional redundant verification can be achieved, effectively avoiding misjudgment by a single model or human rules, and improving robustness and reliability in complex scenarios.

[0080] Based on the above embodiments, the step of calculating the first decision deviation between the first collision dynamics parameter and the second collision dynamics parameter, and the second decision deviation between the first collision dynamics parameter and the third collision dynamics parameter, and determining the automatic emergency braking decision result based on the first decision deviation and / or the second decision deviation, includes: Calculate the first decision bias; If the first decision deviation is less than the first preset threshold, then the first collision dynamics parameter is determined as the automatic emergency braking decision result; If the first decision deviation is greater than or equal to the first preset threshold, then the second decision deviation is calculated; If the second decision deviation is less than the second preset threshold, the first collision dynamics parameter is determined as the automatic emergency braking decision result; if the second decision deviation is greater than or equal to the second preset threshold, the automatic emergency braking decision result is determined to be empty.

[0081] Specifically, the first decision bias and the second decision bias can be calculated sequentially, with the first decision bias calculated first and the second decision bias calculated later. The second decision bias can be calculated only when certain conditions are met.

[0082] After calculating the first decision deviation, if the first decision deviation is less than the first preset threshold, for example, the first preset threshold can be set to 10%, the first decision deviation is the relative deviation between the collision time in the first collision dynamics parameter and the second collision dynamics parameter. If the first decision deviation is less than 10%, it is considered that the outputs of the first automatic emergency braking model and the second automatic emergency braking model are basically consistent. The first automatic emergency braking model has high robustness, so the first collision dynamics parameter can be directly determined as the automatic emergency braking decision result.

[0083] If the first decision deviation is greater than or equal to the first preset threshold, it indicates a significant difference between the outputs of the first and second automatic emergency braking models. In this case, a human-rule decision model is activated to calculate the second decision deviation and determine its relationship with the second preset threshold. If the second decision deviation is less than the second preset threshold, it indicates that the output of the first automatic emergency braking model has passed the human-rule verification, and the first collision dynamics parameter can be directly determined as the automatic emergency braking decision result. If the second decision deviation is greater than or equal to the second preset threshold, it indicates that the output of the first automatic emergency braking model has not passed the human-rule verification, and the output of the first automatic emergency braking model has significant uncertainty. In this case, the automatic emergency braking decision result is determined to be empty, i.e., the automatic braking action is abandoned to prevent false triggering.

[0084] It is understandable that both the first preset threshold and the second preset threshold can be set as needed. They can be the same or different, and no specific restrictions are made here.

[0085] In this embodiment of the invention, by judging conditions at multiple levels, it is determined step by step whether the first collision dynamics parameter can be used as the result of automatic emergency braking. This ensures that the active safety system responds in a timely manner while completely eliminating abnormal braking commands caused by uncertain factors.

[0086] Based on the above embodiments, the step of making a comprehensive decision based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter to determine the automatic emergency braking decision result further includes: The operational status of the first automatic emergency braking model and the second automatic emergency braking model is evaluated in real time. If both the first automatic emergency braking model and the second automatic emergency braking model are operating normally, a comprehensive decision is made based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter to determine the automatic emergency braking decision result.

[0087] Specifically, when making a comprehensive decision using the first, second, and third collision dynamics parameters, the operational status of the first and second automatic emergency braking models can be evaluated in real time. This operational status can be represented by at least one indicator, such as frame rate and data packet loss rate, during the inference calculation process. Here, a pre-set health monitoring module can be used to evaluate the operational status of the first and second automatic emergency braking models in real time by monitoring the real-time data stream.

[0088] If the operation status of the first automatic emergency braking model and the second automatic emergency braking model is normal, for example, if the frame rate of the first automatic emergency braking model and the second automatic emergency braking model are both stable within the preset frame rate range and the data packet loss rate is both stable within the preset packet loss rate range, it indicates that there are no abnormalities in the data processing and transmission process. Then, the first collision dynamics parameter, the second collision dynamics parameter and the third collision dynamics parameter can be used to make a comprehensive decision to determine the automatic emergency braking decision result.

[0089] In this embodiment of the invention, by evaluating the operational status of the first and second automatic emergency braking models in real time, potential single-point failures in hardware or software or network crashes can be detected in a timely manner. This provides a reliable triggering basis for subsequent comprehensive decision-making and ensures system-level functional redundancy. When both the first and second automatic emergency braking models are operating normally, a comprehensive decision is made using three types of collision dynamics parameters. This fully leverages the advantages of multi-model verification, effectively prevents identification or calculation errors from a single network, avoids the system from falsely triggering ghost braking under complex real-world road conditions, and significantly improves the robustness and adaptability of the decision-making algorithm to real-world scenarios.

[0090] Based on the above embodiments, the real-time evaluation of the operational status of the first automatic emergency braking model and the second automatic emergency braking model further includes: If the first automatic emergency braking model is operating normally and the second automatic emergency braking model is operating abnormally, a comprehensive decision is made based on the first collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result. If the first automatic emergency braking model is in an abnormal operating state while the second automatic emergency braking model is in a normal operating state, a comprehensive decision is made based on the second collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result. If the operational states of both the first automatic emergency braking model and the second automatic emergency braking model are abnormal, the automatic emergency braking decision result is determined based on the third collision dynamics parameter.

[0091] Specifically, if the first automatic emergency braking model operates normally, but the second automatic emergency braking model operates abnormally—that is, if the first automatic emergency braking model runs smoothly while the second automatic emergency braking model experiences data packet loss or a sudden drop in frame rate—a degradation mechanism is triggered. In this case, the data verification of the second automatic emergency braking model is automatically skipped, and a comprehensive decision is made directly using the first collision dynamics parameters output by the first automatic emergency braking model and the third collision dynamics parameters output by the human-rule decision-making model. For example, a second decision deviation is calculated and compared with a second preset threshold; the automatic emergency braking decision result is determined based on the comparison result.

[0092] This step, by shielding faulty nodes when the redundant network malfunctions, can prevent the entire decision-making algorithm from being blocked or delayed due to the failure of the secondary automatic emergency braking model, thus ensuring that the core primary automatic emergency braking model can continue to seamlessly guide the automatic emergency braking action.

[0093] If the first automatic emergency braking model malfunctions while the second automatic emergency braking model functions normally (e.g., the first model experiences data interruption or processing errors due to uneven computing power distribution or other reasons, but the second model continues to operate normally), the second collision dynamics parameter output by the second automatic emergency braking model and the third collision dynamics parameter output by the human rule-based decision model are used for collaborative verification to obtain the automatic emergency braking decision result. For example, the third decision deviation between the second and third collision dynamics parameters is calculated and compared with a third preset threshold. If the third decision deviation is less than the third preset threshold, the second collision dynamics parameter can be used as the automatic emergency braking decision result. If the third decision deviation is greater than or equal to the third preset threshold, the third collision dynamics parameter is used as the automatic emergency braking decision result, or the automatic emergency braking decision result is determined to be empty to prevent false triggering.

[0094] This step ensures that even in the event of a localized hardware or software failure, the decision-making entity can still be switched to an independent and redundant second automatic emergency braking model when the first automatic emergency braking model fails, thereby maintaining a high level of active safety protection.

[0095] If both the first and second automatic emergency braking models experience operational anomalies—that is, if both models crash, have extremely low frame rates, or suffer severe data link interruptions—then the automatic emergency braking decision can be directly determined using the third collision dynamics parameter. For example, the third collision dynamics parameter can be directly used as the automatic emergency braking decision, making the decision entirely dependent on classical physical rules such as manually set safety distances, collision times, and steering coordination.

[0096] This step, by degrading to a human rule-based decision-making model when both automatic emergency braking models fail, builds a final, robust line of defense for vehicle safety, eliminating the risk of the automatic emergency braking function completely failing due to the collapse of complex algorithms.

[0097] Based on the above embodiments, the training steps of the first automatic emergency braking model include: The sensor data samples, vehicle state samples, and actual collision dynamics parameters of the sample vehicle are obtained, and the sensor data samples are labeled with environmental perception data tags. The sensor data samples and the vehicle state samples are input into the first initial model to obtain the environmental perception results and the first collision dynamics parameter prediction results output by the first initial model. Based on the difference between the environmental perception result and the environmental perception data label, and the difference between the first collision dynamics parameter prediction result and the actual collision dynamics parameter, the first initial model is iteratively trained to obtain the first automatic emergency braking model.

[0098] Specifically, during the training process of the first automatic emergency braking model, sensor data samples, vehicle state samples, and actual collision dynamics parameters of the sample vehicle can be acquired first. The sensor data samples can include environmental data detected by various sensors installed on the sample vehicle.

[0099] The sensor data sample is labeled with environmental perception data tags. These tags are pre-labeled environmental perception data contained in the sensor data sample, which may include obstacle information, lane line information, and grid occupancy information.

[0100] The vehicle state sample may include at least one of the following: the sample vehicle's speed, acceleration, steering wheel angle, and pose data.

[0101] Actual collision dynamics parameters refer to the actual collision dynamics parameters of the sample vehicle under the scenarios corresponding to the sensor data samples and vehicle state samples. These actual collision dynamics parameters are the actual control strategies to prevent the sample vehicle from colliding with surrounding obstacles, and can include the actual calculated collision time, the actual braking deceleration of the sample vehicle, and the actual grid occupancy of each obstacle in the LiDAR.

[0102] This step, by acquiring sensor data samples containing rich scene information and multi-dimensional real labels as basic training data, can provide a training foundation for the first automatic emergency braking model without information loss, effectively avoiding the information loss problem caused by the traditional single-sensor layer filtering of key targets.

[0103] Subsequently, sensor data samples and vehicle state samples are input into the first initial model. The first initial model uses an end-to-end direct input method to extract features and predict results, and then outputs environmental perception results and first collision dynamics parameter prediction results.

[0104] Subsequently, by utilizing the difference between the environmental perception results and the environmental perception data labels, the environmental perception loss can be calculated. By utilizing the difference between the predicted results of the first collision dynamics parameters and the actual collision dynamics parameters, the first parameter prediction loss can be calculated. By weighted summing of the environmental perception loss and the first parameter prediction loss, the first comprehensive loss can be calculated.

[0105] The first initial model can be iteratively trained using the first comprehensive loss. When the number of iterations reaches a preset number, or the first comprehensive loss converges, the first automatic emergency braking model is obtained. The preset number of iterations can be set as needed and is not specifically limited here.

[0106] In this embodiment of the invention, by using the original sensor data samples as input and combining them with environmental perception data labels and actual collision dynamic parameters to jointly supervise the training of the first initial model, the technical problems of limited information provision and information transmission loss in the traditional perception layer can be effectively solved. This enables the first automatic emergency braking model to extract complex environmental features without loss and with high efficiency, greatly improving the robustness and decision-making efficiency of the automatic emergency braking decision algorithm under complex road conditions.

[0107] Based on the above embodiments, the training steps of the second automatic emergency braking model include: Acquire environmental perception data samples, vehicle state samples, and actual collision dynamics parameters of the sample vehicle; The environmental perception data samples and the vehicle state samples are converted into data vectors, and the data vectors are input into the second initial model to obtain the second collision dynamics parameter prediction results output by the second initial model; Based on the difference between the predicted collision dynamics parameters and the actual collision dynamics parameters, the second initial model is iteratively trained to obtain the second automatic emergency braking model.

[0108] Specifically, during the training process of the second automatic emergency braking model, environmental perception data samples, vehicle state samples, and actual collision dynamics parameters of the sample vehicle can be acquired first. Here, the environmental perception data samples can be either the environmental perception data labels marked in the sensor data samples of the sample vehicle, or the environmental perception data obtained and output by the first automatic emergency braking model using the sensor data samples and vehicle state samples of the sample vehicle as input; no specific limitation is made here.

[0109] Subsequently, the environmental perception data samples and vehicle state samples can be converted into data vectors respectively, and each data vector can be input into the second initial model to obtain the second collision dynamics parameter prediction results output by the second initial model. Alternatively, the environmental perception data samples and vehicle state samples can be stitched together first, and the stitched result can be converted into a data vector.

[0110] Subsequently, the difference between the predicted second collision dynamics parameter and the actual collision dynamics parameter can be calculated. This difference may include at least one of the differences in collision time, steering angle, and braking deceleration. Then, based on the calculated difference, the second parameter prediction loss is calculated.

[0111] The second parameter prediction loss is used as the second comprehensive loss to iteratively train the second initial model. When the number of iterations reaches the preset number, or the second comprehensive loss converges, the second automatic emergency braking model is obtained.

[0112] In this embodiment of the invention, by converting structured environmental perception data samples into data vectors and using them to train an independent second initial model, an independent redundant computing layer based on heterogeneous computing is cleverly constructed. This scheme not only avoids the single-point failure risk of a single deep learning model, but also effectively prevents the false triggering of the automatic emergency braking system through multi-network verification, significantly improving the comprehensiveness and reliability of the autonomous driving active safety system.

[0113] like Figure 3 As shown, this embodiment of the invention also provides an automatic emergency braking control method, which is applied to a target vehicle and includes: S21, Obtain the automatic emergency braking decision result of the target vehicle, wherein the automatic emergency braking decision result is obtained based on the automatic emergency braking decision method provided in the above embodiments; S22, Based on the automatic emergency braking decision result, perform automatic emergency braking control on the target vehicle.

[0114] Specifically, the automatic emergency braking control method provided in this embodiment of the invention is executed by an automatic emergency braking control device, which can be configured in the vehicle controller, autonomous driving domain controller, vehicle computer, or cloud server of the target vehicle. The target vehicle is the vehicle currently in motion that requires automatic emergency braking control.

[0115] First, step S21 is executed, and the automatic emergency braking decision-making method provided in the above embodiments is used to obtain the automatic emergency braking decision result of the target vehicle.

[0116] Then, step S22 is executed, using the automatic emergency braking decision result to generate a control command, and using the control command to perform automatic emergency braking control on the target vehicle, so that the target vehicle brakes according to the braking deceleration in the automatic emergency braking decision result or steers according to the steering angle in the automatic emergency braking decision result, so as to avoid collision with surrounding obstacles.

[0117] The automatic emergency braking control method provided in this embodiment of the invention utilizes the automatic emergency braking decision results obtained from comprehensive decision-making to achieve automatic emergency braking control. This can improve the response sensitivity to complex and ever-changing actual road conditions, avoid false triggering behaviors such as ghost braking, comprehensively enhance the robustness and reliability of the target vehicle's active safety system, and reduce the probability of vehicle accidents.

[0118] like Figure 4 As shown, based on the above embodiments, the automatic emergency braking decision-making method and automatic emergency braking control method provided in this embodiment of the invention include: Acquire the sensor data and vehicle status data (ego_info) of the target vehicle. Sensor data and vehicle status data are input into the first automatic emergency braking model to obtain the obstacle detection results, lane line detection results, occupancy grid detection results, first braking deceleration, first collision time and first steering angle output by the first automatic emergency braking model; The obstacle detection results, lane line detection results, occupancy grid detection results and vehicle status data output by the first automatic emergency braking model are converted into data vectors and then input into the second automatic emergency braking model to obtain the second braking deceleration, second collision time and second steering angle output by the second automatic emergency braking model. The obstacle detection results, lane line detection results, occupancy grid detection results and vehicle status data output by the first automatic emergency braking model are converted into data vectors and then input into the human rule decision model to obtain the third braking deceleration, third collision time and third steering angle output by the human rule decision model. By comprehensively considering the first braking deceleration, the first collision time, the first steering angle, the second braking deceleration, the second collision time, the second steering angle, the third braking deceleration, the third collision time, and the third steering angle, an automatic emergency braking decision is obtained. The automatic emergency braking decision results are used to control the target vehicle to perform automatic emergency braking.

[0119] like Figure 5 As shown, based on the above embodiments, this embodiment of the invention provides an automatic emergency braking decision device, applied to a target vehicle, comprising: The data acquisition module 41 is used to acquire sensor data and vehicle status data of the target vehicle; The first decision module 42 is used to input the sensor data and the vehicle status data into the first automatic emergency braking model to obtain the environmental perception data and the first collision dynamics parameters output by the first automatic emergency braking model. The second decision module 43 is used to input the environmental perception data and the vehicle status data into the second automatic emergency braking model to obtain the second collision dynamics parameters output by the second automatic emergency braking model. The third decision module 44 is used to input the environmental perception data and the vehicle state data into the human rule decision model to obtain the third collision dynamics parameters output by the human rule decision model. The integrated decision module 45 is used to make an integrated decision based on the first collision dynamics parameter, the second collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result.

[0120] Specifically, the functions of each module in the automatic emergency braking decision device provided in this embodiment of the invention correspond one-to-one with the operation flow of each step in the above method-like embodiments, and the achieved effects are also the same. For details, please refer to the above embodiments, and this will not be repeated in this embodiment of the invention.

[0121] like Figure 6 As shown, based on the above embodiments, this embodiment of the invention provides an automatic emergency braking control device applied to a target vehicle, comprising: The decision result acquisition module 51 is used to acquire the automatic emergency braking decision result of the target vehicle, and the automatic emergency braking decision result is obtained based on the automatic emergency braking decision method provided in the above embodiments. The brake control module 52 is used to perform automatic emergency braking control on the target vehicle based on the automatic emergency braking decision result.

[0122] Specifically, the functions of each module in the automatic emergency braking control device provided in the embodiments of the present invention correspond one-to-one with the operation flow of each step in the above method embodiments, and the achieved effects are also the same. For details, please refer to the above embodiments, and this will not be repeated in the embodiments of the present invention.

[0123] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7 As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute the automatic emergency braking decision method or the automatic emergency braking control method provided in the above embodiments.

[0124] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to related technologies, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the automatic emergency braking decision method or the automatic emergency braking control method provided in the above embodiments.

[0126] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the automatic emergency braking decision-making method or the automatic emergency braking control method provided in the above embodiments. This computer-readable storage medium can be either a non-transitory computer-readable storage medium or a transient computer-readable storage medium, and is not specifically limited herein.

[0127] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0128] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of software products. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An automatic emergency braking decision-making method, characterized in that, Applied to target vehicles, including: Acquire sensor data and vehicle status data of the target vehicle; The sensor data and the vehicle status data are input into the first automatic emergency braking model to obtain the environmental perception data and the first collision dynamics parameters output by the first automatic emergency braking model. The environmental perception data and the vehicle status data are input into the second automatic emergency braking model to obtain the second collision dynamics parameters output by the second automatic emergency braking model. The environmental perception data and the vehicle state data are input into the human rule decision model to obtain the third collision dynamics parameter output by the human rule decision model; The automatic emergency braking decision is determined by comprehensively considering the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter.

2. The automatic emergency braking decision-making method according to claim 1, characterized in that, The step of making a comprehensive decision based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter to determine the automatic emergency braking decision includes: Based on the first collision dynamics parameters, the risk level of the current scenario of the target vehicle is classified. If the result obtained from the classification is the target level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the first collision dynamics parameter is verified to determine the automatic emergency braking decision result; If the result of the classification is not the target level, then the automatic emergency braking decision result is determined based on the first collision dynamics parameter.

3. The automatic emergency braking decision-making method according to claim 2, characterized in that, The target level includes a warning level and an emergency level; the first collision dynamics parameters include collision time, braking deceleration, and steering angle. The step of verifying the first collision dynamics parameter based on the second and third collision dynamics parameters to determine the automatic emergency braking decision result includes: If the result obtained by the grading is the warning level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the collision time and braking deceleration in the first collision dynamics parameter are verified to determine the automatic emergency braking decision result; If the result obtained from the classification is the emergency level, then based on the second collision dynamics parameter and the third collision dynamics parameter, the collision time, braking deceleration and steering angle in the first collision dynamics parameter are verified to determine the automatic emergency braking decision result.

4. The automatic emergency braking decision-making method according to claim 2, characterized in that, The step of verifying the first collision dynamics parameter based on the second and third collision dynamics parameters to determine the automatic emergency braking decision result further includes: Calculate the first decision deviation between the first collision dynamics parameter and the second collision dynamics parameter, and the second decision deviation between the first collision dynamics parameter and the third collision dynamics parameter, and determine the automatic emergency braking decision result based on the first decision deviation and / or the second decision deviation.

5. The automatic emergency braking decision-making method according to claim 4, characterized in that, The step of calculating a first decision deviation between the first collision dynamics parameter and the second collision dynamics parameter, and a second decision deviation between the first collision dynamics parameter and the third collision dynamics parameter, and determining the automatic emergency braking decision result based on the first decision deviation and / or the second decision deviation, includes: Calculate the first decision bias; If the first decision deviation is less than the first preset threshold, then the first collision dynamics parameter is determined as the automatic emergency braking decision result; If the first decision deviation is greater than or equal to the first preset threshold, then the second decision deviation is calculated; If the second decision deviation is less than the second preset threshold, the first collision dynamics parameter is determined as the automatic emergency braking decision result; if the second decision deviation is greater than or equal to the second preset threshold, the automatic emergency braking decision result is determined to be empty.

6. The automatic emergency braking decision-making method according to any one of claims 1-5, characterized in that, The step of making a comprehensive decision based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter to determine the automatic emergency braking decision result further includes: The operational status of the first automatic emergency braking model and the second automatic emergency braking model is evaluated in real time. If both the first automatic emergency braking model and the second automatic emergency braking model are operating normally, a comprehensive decision is made based on the first collision dynamics parameter, the second collision dynamics parameter, and the third collision dynamics parameter to determine the automatic emergency braking decision result.

7. The automatic emergency braking decision-making method according to claim 6, characterized in that, The real-time evaluation of the operational status of the first automatic emergency braking model and the second automatic emergency braking model further includes: If the first automatic emergency braking model is operating normally and the second automatic emergency braking model is operating abnormally, a comprehensive decision is made based on the first collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result. If the first automatic emergency braking model is in an abnormal operating state while the second automatic emergency braking model is in a normal operating state, a comprehensive decision is made based on the second collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result. If the operational states of both the first automatic emergency braking model and the second automatic emergency braking model are abnormal, the automatic emergency braking decision result is determined based on the third collision dynamics parameter.

8. The automatic emergency braking decision-making method according to any one of claims 1-5, characterized in that, The training steps for the first automatic emergency braking model include: The sensor data samples, vehicle state samples, and actual collision dynamics parameters of the sample vehicle are obtained, and the sensor data samples are labeled with environmental perception data tags. The sensor data samples and the vehicle state samples are input into the first initial model to obtain the environmental perception results and the first collision dynamics parameter prediction results output by the first initial model. Based on the difference between the environmental perception result and the environmental perception data label, and the difference between the first collision dynamics parameter prediction result and the actual collision dynamics parameter, the first initial model is iteratively trained to obtain the first automatic emergency braking model.

9. The automatic emergency braking decision-making method according to any one of claims 1-5, characterized in that, The training steps for the second automatic emergency braking model include: Acquire environmental perception data samples, vehicle state samples, and actual collision dynamics parameters of the sample vehicle; The environmental perception data samples and the vehicle state samples are converted into data vectors, and the data vectors are input into the second initial model to obtain the second collision dynamics parameter prediction results output by the second initial model; Based on the difference between the predicted collision dynamics parameters and the actual collision dynamics parameters, the second initial model is iteratively trained to obtain the second automatic emergency braking model.

10. An automatic emergency braking control method, characterized in that, Applied to target vehicles, including: The automatic emergency braking decision result of the target vehicle is obtained, and the automatic emergency braking decision result is obtained based on the automatic emergency braking decision method as described in any one of claims 1-9; Based on the automatic emergency braking decision results, automatic emergency braking control is performed on the target vehicle.

11. An automatic emergency braking decision-making device, characterized in that, Applied to target vehicles, including: The data acquisition module is used to acquire sensor data and vehicle status data of the target vehicle; The first decision module is used to input the sensor data and the vehicle status data into the first automatic emergency braking model to obtain the environmental perception data and the first collision dynamics parameters output by the first automatic emergency braking model. The second decision module is used to input the environmental perception data and the vehicle status data into the second automatic emergency braking model to obtain the second collision dynamics parameters output by the second automatic emergency braking model. The third decision module is used to input the environmental perception data and the vehicle state data into the human rule decision model to obtain the third collision dynamics parameters output by the human rule decision model. The integrated decision module is used to make an integrated decision based on the first collision dynamics parameter, the second collision dynamics parameter and the third collision dynamics parameter to determine the automatic emergency braking decision result.

12. An automatic emergency braking control device, characterized in that, Applied to target vehicles, including: The decision result acquisition module is used to acquire the automatic emergency braking decision result of the target vehicle, wherein the automatic emergency braking decision result is obtained based on the automatic emergency braking decision method as described in any one of claims 1-9; The brake control module is used to perform automatic emergency braking control on the target vehicle based on the automatic emergency braking decision result.

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the automatic emergency braking decision method as described in any one of claims 1-9, or the automatic emergency braking control method as described in claim 10.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the automatic emergency braking decision method as described in any one of claims 1-9, or the automatic emergency braking control method as described in claim 10.

Images